Infrared light-responsive CuSbS2 optoelectronic artificial synapses enabling high-accuracy color image recognition and classification

Abstract

Near-infrared (NIR) optoelectronic synaptic memristors, functioning as devices capable of realizing parallel information perception and memory, hold significant importance for the development of high-efficiency neuromorphic computing systems that integrate visual perception and complex learning capabilities. Herein, CuSbS2 is innovatively introduced to design a NIR-responsive artificial optoelectronic synapse device. Such CuSbS2-based memristors demonstrate a stable non-volatile bipolar resistive switching behavior, with an on/off ratio of ∼99 and excellent cycling endurance exceeding 104 s. The device can be triggered by electrical signals and NIR light stimuli to emulate various synaptic plasticity behaviors, including excitatory postsynaptic current, paired-pulse facilitation, short-term potentiation/depression, spike-timing/voltage/duration/rate-dependent plasticity (ST/V/D/RDP), and the “learning–consolidation” behavior. Notably, the CuSbS2 memristor-based neuromorphic computing system achieves a high accuracy of 92.3% in color image recognition and classification tasks. This indicates that the learning rate regulation mechanism supported by the device can significantly improve the training efficiency and model generalization capability, highlighting its potential in intelligent visual perception applications. These findings confirm that CuSbS2 is a highly promising NIR optoelectronic artificial synaptic functional material, which is expected to advance the integration of sensing, storage, and information processing, offering a novel material platform for future brain-inspired computing and artificial intelligence systems.